The Apple A6 SOC: Cortex A15 Hits The Market

Subject: Processors

Apple Produces the new A6 for the iPhone 5


Today is the day that world gets introduced to the iPhone 5.  I of course was very curious about what Apple would be bringing to market the year after the death of Steve Jobs.  The excitement leading up to the iPhone announcement was somewhat muted as compared to years past, and a lot of that could be attributed to what has been happening in the Android market.  Companies like Samsung and HTC have released new high end phones that are not only faster and more expansive than previous versions, but they also worked really well and were feature packed.  While the iPhone 5 will be another success for Apple, for those somewhat dispassionate about the cellphone market will likely just shrug and say to themselves, “It looks like Apple caught up for the year, but too bad they really didn’t introduce anything really groundbreaking.”

View Full Size

If there was one area that many were anxiously awaiting, it was that of the SOC (system on a chip) that Apple would use for the iPhone 5.  Speculation went basically from using a fresh piece of silicon based on the A5X (faster clocks, smaller graphics portion) to having a quad core monster running at high speeds but still sipping power.  It seems that we actually got something in between.  This is not a bad thing, but as we go forward we will likely see that the silicon again only matches what other manufacturers have been using since earlier this year.

Click here to read the entire article.

Apple was very limited on what they said about the A6, and the only real details we have is that it is a dual core processor based on ARM Cortex A15 technology.  Apple had bought up a ARM development group some years back and have been using their in-house designs for SOCs since that point.  Apple still does not own any Fabs (nor do they really want to), so they have to go to another manufacturer to actually produce these chips.  Previously Apple has used Samsung as their primary foundry partner, and even though those two companies are involved in some very bitter patent disputes, the relationship with the foundry side of Samsung and Apple has never been contentious.

Previous chips such as the A5 and A5X were produced on Samsungs 45 nm process.  Competing SOCs such as NVIDIA’s Tegra 2 and Tegra 3 are produced on TSMC’s 40 nm process.  This year Qualcomm released their latest Snapdragon 4 variants using TSMC’s 28 nm process.  The jump from 45 nm to 28 nm is pretty significant, and plays a big factor in what Apple seems to have done.  The choice of design also plays a big factor, as the Cortex A9 is actually getting a bit long in the tooth.

From what we have gathered, it appears as though the A6 is a 32 nm part fabricated by Samsung.  There is no evidence that Apple has actually contracted out TSMC as a 28 nm partner, but there has been talk of Apple looking at TSMC as a second source.  When we consider the timeline of SOC development, it would take quite a bit of extra time to move designs to another foundry partner with a radically different process technology.  Apple said that this particular part is twice as fast as the previous, and when considering design and potential clockspeeds, this is a not a specious proclamation.

View Full Size

The Apple A6 actually appears to be the first “official” Cortex A15 SOC in mass production.  Samsung Exynos 4 Dual and Quads are based on Cortex A9, and the upcoming Exynos 5 Dual is Cortex A15 but not released as of yet.  Qualcomm has a hybrid style design that is entirely homebrewed, and it can be viewed as a Cortex A9+ which mirrors some processing features of the A15.  Codenamed Krait, this product is more advanced than an A9, but it is not completely an A15 SOC.  The Cortex A15 is a more powerful Cortex variant that can achieve higher throughput and clockspeeds than previous A series chips.  More power saving features have been implemented in the design, and it does increase the IPC of the Cortex series at the same time.  This makes the processor more efficient overall, and with a good combination of firmware and process technology can create a nice leap in mobile performance.  We have seen this with Qualcomm’s S4 MSM8960, which powers competing products like the Samsung Galaxy S3, the HTC One-X, the Nokia Lumia 920, and the Motorola RAZR HD.

If our assumption is correct that this particular product is based on the Samsung 32 nm process, then we can safely assume that the A6 in the iPhone 5 will be clocked in the 1.3 GHz to 1.7 GHz range.  Considering that the A5 was a 45 nm dual core part based on the Cortex A9, and that ran at 800 MHz, a conservative call of the A6 running at 1.5 GHz is pretty acceptable.  At such a speed it will not pull any more power than the previous generation A5 when at full load.  In this case it will be running significantly faster in both clockspeed and measured performance.  Chances are that at 1.5 GHz the A6 will slightly outperform the Krait powered Snapdragon 4 series of chips which are again generally clocked at 1.5 GHz in the above listed phones.  At the very least it performs at near the same level.  This is not Apple’s first attempt at a 32 nm part, the revised A5 used in the iPad 2 (second version) is a 32 nm part and a nice stepping stone to get acquainted with Samsung’s (then) new process line.

The Cortex A15 is specified to go upwards of 2 GHz in certain applications, such as larger form factor notes and tablets that will have better thermal dissipation characteristics as well as larger batteries.  Apple wants to achieve acceptable battery life, yet still perform as well as the rest of the crowd.  Underclocking the A15 based A6 to 1.4 or 1.5 GHz should achieve very favorable results in this particular application.  Electrically speaking, Samsung’s 32 nm process has many of the same leakage and transistor switching properties as TSMC’s 28 nm bulk process.  The only area that the TSMC process is better in is that of transistor density.  If Apple were to jump ship at Samsung and go for TSMC, they would have many of the same power and speed characteristics, but they would be able to produce a smaller die and fit more chips on each wafer.

Apple made absolutely no mention of the graphics portion of the A6 other than it was faster than the previous generation.  Apple has been licensing Imagination Technologies Power VR products since the original iPhone.  The latest iPad uses the A5X SOC, which features the SGX543MP4 (quad core graphics).  The PowerVR Series 6 “Rogue” graphics have not been integrated into a SOC yet, and the first examples are expected in Q1 2013.  This means that the likely candidate for the iPhone 5 and the A6 SOC is the SGX543MP4.  This will achieve more than double the graphics performance of the previous iPhone 4S, not only due to the greater amount of shaders present but also to the faster clockspeed that the graphics core will run at on Samsung’s 32 nm process.  On the other hand the SGX544MP2 looks to be showing up in the updated Intel Medfield SOC.  The 543 was originally released in 2009, and the updated 544 was released in 2010.  It makes sense that Apple would continue to utilize the PowerVR licenses as their entire previous code base is tied directly to those products.

This is about the extent of what I could come up with considering the limited information that Apple has given.  We have to drag clues from all over the industry to come to these suppositions, and until we get official word from Apple we will not know for sure.  Do not take this as gospel, but I do believe that this is a good start.

Video News

September 12, 2012 | 09:47 PM - Posted by Anonymous (not verified)

Thanks for giving some idea of actual performance, when companies give vague statements I get annoyed, like they're dodging reality, twice as fast to me sounds like they found one thing the processor does twice as fast and the rest is only marginally better but numbers never lie.

September 13, 2012 | 01:38 AM - Posted by Hazydave (not verified)

Apple said "smaller chip", in the context of making a smaller PCB, so they could stuff in more battery. They didn't actually claim the A6 die was smaller than the A5 die. However, the extra speed and presence of what looks like a dual core A15 processor pretty much ensures they're using Samsung's 32nm process for this, rather than the 45nm process used on the A5.

September 20, 2012 | 04:24 AM - Posted by Anonymous (not verified)

Since TSMC has turned down both Apple's and Qualcomm's approaches for their capacity, Apple has no other choice in foundry partners unless they can get Intel to do such jobs for them. So they are stuck with Samsung's 32nm node for some time yet.
The A6 is not following the A15 design but just improvements on the memory controller (LP-ddr2 1066Mhz) and bigger cache (1024K vs 256KB) and other smaller improvements to the A9. It would be a A9 variant rather than A15 since the A15 has a longer pipeline . We can only know this if Apple ever disclose the architecture of the A6.

September 20, 2012 | 04:24 AM - Posted by Anonymous (not verified)

Since TSMC has turned down both Apple's and Qualcomm's approaches for their capacity, Apple has no other choice in foundry partners unless they can get Intel to do such jobs for them. So they are stuck with Samsung's 32nm node for some time yet.
The A6 is not following the A15 design but just improvements on the memory controller (LP-ddr2 1066Mhz) and bigger cache (1024K vs 256KB) and other smaller improvements to the A9. It would be a A9 variant rather than A15 since the A15 has a longer pipeline . We can only know this if Apple ever disclose the architecture of the A6.

September 12, 2012 | 10:40 PM - Posted by Susanna (not verified)

SGX543MP4 would not be smaller as apple has stated, even with the die shrink.

SGX544MP2 does not bring extra performance, only support for certain Direct X transformations.

Apple has been first with PowerVR Cores before (they are investor, why wouldn't they be first?)

September 13, 2012 | 12:29 AM - Posted by Josh Walrath

Why would SGX543MP4 not be smaller at 32 nm than at 45 nm?

September 14, 2012 | 01:12 AM - Posted by dagamer34 (not verified)

I think he meant a A6 with a MP4 would not be smaller than an A5 with an MP2. The GPU cores take up a lot of die space.

September 15, 2012 | 01:15 AM - Posted by Anonymous (not verified)

1.6ghz CPU sgx543mp4 and 2gb on tmsc 28nm or 32nm with 2gb of ram would be nice.

September 14, 2012 | 01:15 AM - Posted by dagamer34 (not verified)

Keep in mind that clock speed can change affecting performance.

September 13, 2012 | 02:23 AM - Posted by Anonymous (not verified)

Good insight.
Small correction, the HTC One-X uses Tegra 3 not the S4 MSM8960.

September 13, 2012 | 03:15 AM - Posted by Anonymous (not verified)

I think he's talking about the US variants since he mention the Samsung Galaxy S3 and others, which are also not using Qualcomm's chip internationally, but use it in their US variants.

September 13, 2012 | 03:15 AM - Posted by JRod (not verified)

This seems like what should have been used on the new iPad.

September 13, 2012 | 10:24 AM - Posted by Darkseider (not verified)

All of this is purely speculative at the moment. It is either a dual core Cortex A15 or a higher clocked dual core Cortex A9 built using the smaller process.

September 13, 2012 | 10:55 AM - Posted by Josh Walrath

Except in the presentation he did mention it was a dual core Cortex A15 processor.

September 13, 2012 | 11:58 AM - Posted by Anonymous (not verified)

I don't think it was mentioned.

September 13, 2012 | 01:28 PM - Posted by Activate: AMD (not verified)

I haven't seen anywhere that has definitively proven its an A15, just analyst speculation or guessing.

I have to admit, if its A15 I'm surprised. Lots of other SoC manufacturers out there haven't even demo'd working hardware with A15, and now Apple is claiming to be shipping in massive volume? I know they have huge resources, but to have A15 cores that far ahead of TI or Samsung's own SoC's seems unusual. Either way, Android has had A15-like cores for the past 3-4 months in the form of Krait.

September 13, 2012 | 04:15 PM - Posted by Josh Walrath

We tracked down where the A15 information was pushed out.  Anand's liveblog of the event.  Right as they introduced the hardware behind it, Anand said it was a dual core A15.  I think this is where a lot of people got the A15 information from.  I'm not sure why he said it, but apparently a ex-system engineer from Apple told him about it (Jim Keller perhaps?).

September 14, 2012 | 11:10 AM - Posted by Activate: AMD (not verified)

From what I could see Anand basically said on his site that apple reserves generational CPU numbers for generational Changes (i.e. A4->A5 = Cortex A8 -> Cortex A9). If thats the basis for the information being thrown around everywhere, forgive me for wanting a little more confirmation than that. Realistically, hes probably right, but I am quite surprised that they were able to get A15 into a product at this date

September 13, 2012 | 01:46 PM - Posted by Anonymous (not verified)

Apple is just copying android

September 17, 2012 | 07:51 AM - Posted by Anonymous (not verified)


September 18, 2012 | 01:07 AM - Posted by Anonymous (not verified)

The question is who is going to make it now neither Apple or Samsung are not interested in renewing their current contract. TSMC were not fans of Apples demands for fab exclusivity and are already at 90% production capacity with Nvidia and Qualcomm orders.

September 19, 2012 | 12:10 PM - Posted by Josh Walrath

Samsung is a huge company.  As such, apparently certain parts do not necessarily agree or get along all that well with other parts.  The fab guys at Samsung are trying to distance themselves from the phone guys, and apparently the fab arm of Samsung has a very good relationship with Apple.  So, I don't see Apple moving away from Samsung anytime soon when it comes to fabricating chips.  And yes, you are correct, TSMC does not have the 28 nm capacity to handle Apple right now (or their exclusivity demands).

January 4, 2013 | 07:23 AM - Posted by Anonymous (not verified)

It took three months..
This it not official yet but the papers seem to be signed?

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

This question is for testing whether you are a human visitor and to prevent automated spam submissions.